Cleaned_up_regression

Using distance traveled as dependent variable

Visualize data

Fuel prices

Seasonal decomposition of distance traveled time series from 2013 to 2016

1. Monthly time steps

Model specifications:

Model 1.1: Just log-log

\[ ln(\text{distance_traveled}) = \alpha_0 + \alpha_1 ln(\text{fuel_price}) + \epsilon\]

Model 1.2: Monthly Log-Log without lags

\[ ln(\text{distance_traveled}) = \alpha_0 + \alpha_1 ln(\text{fuel_price}) + \alpha_2 \text{country} + \alpha_3 \text{month} + \epsilon\]

Model 1.3: Monthly log-log with lagged fuel price

\[ ln(\text{distance_traveled}) = \alpha_0 + \alpha_1 ln(\text{fuel_price}_{t-1}) + \alpha_2 \text{country} + \alpha_3 \text{month} + \epsilon \]

Models summary

	Dependent variable:
	log(distance_traveled)
	(1)	(2)	(3)
log(fuel_price)	-0.031***	-0.022***
	(-0.045, -0.016)	(-0.034, -0.009)
log(lagged_fuel_price)			-0.021***
			(-0.034, -0.008)
Constant	13.707***	12.597***	12.595***
	(13.612, 13.803)	(12.512, 12.683)	(12.510, 12.680)
Observations	349,676	349,676	342,990
R2	0.00005	0.256	0.255
Adjusted R2	0.00004	0.256	0.254
Residual Std. Error	1.492 (df = 349674)	1.287 (df = 349586)	1.285 (df = 342900)
F Statistic	16.651*** (df = 1; 349674)	1,351.158*** (df = 89; 349586)	1,315.675*** (df = 89; 342900)
Note:	p<0.1; p<0.05; p<0.01

Comparing elasticity estimates with other published estimates in other sectors

2. Weekly time steps

Model specifications:

model 2.1: Just log-log:

\[ ln(\text{distance_traveled}) = \alpha_0 + \alpha_1 ln(\text{fuel_price})\]

model 1.2: Log-Log without lags:

\[ ln(\text{distance_traveled}) = \alpha_0 + \alpha_1 ln(\text{fuel_price}) + \alpha_2 \text{country} + \alpha_3 \text{month} + \alpha_4\text{new_years} + \epsilon \]

model 2.3. Log-Log with lagged fuel price:

\[ ln(\text{distance_traveled}) = \alpha_0 + \alpha_1 ln(\text{fuel_price}_{t-1}) + \alpha_2 \text{country} + \alpha_3 \text{month} + \alpha_4\text{new_years} + \alpha_5\text{distance_traveled}_{t-1} + \epsilon \]

Summary of results

	Dependent variable:
	log(distance_traveled)
	(1)	(2)	(3)
log(fuel_price)	-0.065***	-0.063***
	(-0.072, -0.058)	(-0.070, -0.056)
new_year		-0.129***	-0.130***
		(-0.151, -0.107)	(-0.152, -0.108)
log(lagged_fuel_price)			-0.064***
			(-0.070, -0.057)
Constant	12.912***	12.201***	12.206***
	(12.865, 12.959)	(11.960, 12.442)	(11.964, 12.447)
Observations	1,132,619	1,089,051	1,089,051
R2	0.0003	0.206	0.206
Adjusted R2	0.0003	0.206	0.206
Residual Std. Error	1.304 (df = 1132617)	1.156 (df = 1088961)	1.156 (df = 1088961)
F Statistic	315.570*** (df = 1; 1132617)	3,177.313*** (df = 89; 1088961)	3,177.471*** (df = 89; 1088961)
Note:	p<0.1; p<0.05; p<0.01

Comparing elasticity estimates with other published estimates in other sectors

Figure X: Comparison of short-run price elasticity of fuel demand of the global fishing fleet with previous estimates from other sectors. In their meta-analysis, Brons et al, (2008) estimate a short-run price elasticity of gasoline demand of −0.34. Hughes et al, (2008) report a very similar elasticity for the period between 1975 and 1980 (-0.335) but they highlight a sharp decrease to -0.041 in more recent years (2001-2006). More recently, Havnarek et al, (2012) suggest that previous estimates of price elasticites suffer from publication bias and their meta-analysis estimate that after proper correction short-run price elasticity is only -0.09. Lastly, Winebreak et al, (2015) estimate the price elasticity of the U.S combination trucking sector between 1980 and 2012 to be 0.0005 suggesting that this sector is virtually inelastic to changes in fuel price.